sovereign-agent

Sovereign Agent documentation

Sovereign Agent 1.x is an always-on agent you can read end to end: how an outcome becomes governed work performed by accountable actors. It is a compact Python reference implementation, not the production Zero Employee control plane.

Start here

1. Install and run the quickstart.
2. Build it step by step with the book and its exercises.
3. Read the architecture when you want to connect a chapter concept to the production modules.
4. Use the API reference and compatibility policy when integrating the package.

The default curriculum is offline and deterministic. Live provider CLIs are optional and credentialed execution is never required by the default test suite.

Current 1.x reference

• Architecture
• Public API and stability contract
• Task-oriented API reference
• Persistence boundary
• Durable non-goals
• Security policy and threat model
• Migration from 0.7 to 1.x
• Rulings index

Historical 0.x documentation

Files named v0.*, migration-v0.*, the legacy tutorials, generated-reference stubs, and release notes describe the retired pre-1.0 framework. They remain in the repository so old releases and migration decisions stay auditable; they are not a guide to the code on main. See Historical documentation before following one of those pages.

Repository map

Path	Purpose
src/sovereign_agent/	The small governed-organization implementation and CLI.
src/reference_organizations/store/	Lucy’s store domain used by the book.
tests/	Behavioral and adversarial proof matrix.
scripts/	Release, curriculum, proof-pack, and repository verifiers.
docs/	Current reference plus explicitly retained historical records.
.github/	Provider-neutral GitHub automation and contributor templates.

Local editor, model-provider, credential, build, and session state does not belong in the public source tree.

This site is open source. Improve this page.